Revolutionizing Recommender Systems With GANs, LLMs, and Knowledge Graphs

Abstract

Background Information: Traditional recommendation systems suffer from low context-awareness, scarcity of data, and cold start problems. KGs and recent AI techniques enhance recommendations by infusing semantic relationships and contextual knowledge. Objectives: The work will address the limitations of the recommender systems by integrating GANs, LLMs, and KGs in a hybrid framework that brings context awareness, specificity, and diversity. Methods: The system, for semantic understanding and context awareness, uses GANs to enhance data variety and LLMs and KGs. Metrics used to evaluate it against benchmark datasets included precision, recall, diversity, cold-start accuracy, and response time. Results: The hybrid framework outperformed traditional systems with a response time of 183.5 ms, 88.7% precision, 87.3% recall, 81.6% diversity, and 84.3% cold-start accuracy. Conclusion: This approach combines GANs, LLMs, and KGs to provide context-aware, varied, and personalized recommendations, generating a robust, dynamic, and user-centric solution.